Two owners, and therefore none

Four thousand two hundred metres of the wrong blue

On 9 February, Bardawil Mills delivered 4,200 m of shell twill for Halverton's WB-2611. It was dyed to the shade standard Tanvara had sent. That standard was last season's. The cloth was right. The mill was right. The paperwork matched. And the colour was wrong for the garment it was going into.

The cloth itself was worth 20,370.00 and was saved. That is the only reason this is a lesson and not a disaster.

Now the part that matters. Nobody failed to do their job. Two people did the same job. Each believed the other had already done it. Between them, neither did it.

Rasheeda Serageldin runs product development. She owns shade approval: she decides when a lab dip passes. Hanadi Naguib owns the commercial relationship with the mill: she sends the mill what it needs. The standard for WB-2611 was approved on 21 January. Rasheeda assumed it went out with the fabric order, because it usually does. Hanadi assumed it went out with the approval, because it usually does. Both assumptions describe how that office really works. Neither is a mistake in the ordinary sense.

A shared record fails at the handover, not at the keyboard

Course 10.1 measured this for a single field. It found that an unowned field fails far more often than an owned one. For a field, that conclusion is right: two people writing to one field is usually two fields wearing one name, and the fix is a definition.

A record is different, and that is why this is a separate lesson. A style record is not one value with two claimants. It is forty or fifty values moving through five stages. Shared ownership does not produce an argument over a value. It produces an empty seam. Nobody typed a wrong shade standard anywhere. What never happened was a handover between two stages, and a handover has no field.

Hanadi measured the thirty-four styles.

A shared record raised 4.4 times as many exceptions as a singly owned one. The ratio is the less interesting half. Read the eleven exceptions and every one has the same shape: something had to move from one stage to the next, it did not move, and somebody else noticed days later — the mill, the buyer, the trim supplier, the store.

This is why a data-quality audit will never find a shared record. Every value is there. Every value looks sensible. The defect is an event that did not happen, and you cannot write a query for an event that did not happen.

The test that tells a coordinator from an owner

Ask this of any record: if it is wrong for a month, whose week gets worse?

If you can name one person, the record is owned. If you can name nobody, course 10.1 has already priced that for you. If you can name two, you have this lesson's problem. The giveaway is that both of them will describe their job using the word and — I do shade approval and I keep an eye on the mill file. Keeping an eye is not ownership. It is what people do instead of ownership when nobody has assigned it.

The tempting fix is to name one of them the owner of the whole record. It fails for a reason that is easy to say and hard to accept: a style record really has no single owner across its life. In development, product development is accountable and merchandising is a passenger. In production, it is the other way round. Make Rasheeda the owner of WB-2611 for its whole life, and you make her accountable for a shipping mark she has never seen.

Ownership moves, and the handover is an event with a name

What Tanvara did instead cost four hours and a printed table on the wall.

Three things make this work, and you need all three.

1. Exactly one name in each row. Not a department. A person, with a deputy named separately for when they are away.
2. The handover is an event, not a feeling. It has a name, a date, and somebody who performs it. A record that has moved to Confirmed has been handed over. A record that is about to move has not, and its old owner still holds it.
3. Use the events that already exist. Costing signed and fabric booked were already moments people recognised. Inventing five new ceremonies would have produced a procedure that lasted a fortnight, which is what lesson 8 is about.

The following quarter, the thirteen records that used to be shared raised three exceptions instead of eleven: 0.23 per record against 0.85, a fall of 73%. The change cost four hours of meeting time: USD 56.80.

The five other exceptions, which are cheaper and teach more

The cloth was the expensive one. The other five in the ownership row were: a lab dip submitted and paid for twice (370.00), a size set couriered twice (256.80), a swatch card made twice (96.00), a costing revised twice from different assumptions (63.90 of time), and one approval asked of Halverton twice, which cost nothing but a slightly cooler email from Orla Wrenshall.

Notice that four of the five are duplicates, not gaps. That is the second sign of a shared record, and it is the one you can find without waiting for an accident: look for the things your factory does twice. A duplicated action is two owners both being conscientious. That is far easier to spot in a run of invoices than a missing handover is to spot anywhere.

Help me replace shared ownership of a record with one responsible owner for each stage of its life.

Here is the record type, and the stages it passes through: [PASTE]. Here are the people and roles who touch it: [PASTE]. Here are the last ten exceptions raised on records of this type, and what went wrong in each one: [PASTE].

1. For each stage, name exactly ONE responsible person. Use this test: if the record stays wrong for a month at this stage, whose week gets worse? Name a deputy separately.
2. For each move from one stage to the next, name the handover event. Use an event that already happens in our work: a costing signed, a fabric booked, an inspection passed. Never invent a new ceremony.
3. Sort my ten exceptions into two piles. Pile one: things that did not happen at a handover. Pile two: things that happened twice. Tell me which pile is bigger. Things done twice are the cheap early warning. Things not done at all are the expensive one.
4. Point out any stage where the owner I chose cannot actually make the values true. Then tell me who can.

Be blunt if a stage really has no owner today.